The Research Of Registration Algorithm In Augmented Reality

The augmented reality is a new area of research. In an augmented reality system, a user's view of the real world is enhanced or augmented with graphics, text notes and other virtual information generated from a computer.Registration is the critical technology for augmented reality system, it solves the problem of real-time detection of a user's head position and orientation. According to the information determines the mapping position of the additional virtual information, and real time display the virtual information into the correct position of HMD display. Catching the position and direction is the task of tracking algorithm. In recent years, the method of tracking based on computer vision is paid more attention to. The tracking method based on the visual can be applied to more complex environment, and can effectively deal with occlusion, changes in illumination, mistracking., and other issues.AI Comport, D. Kragic, E. Marchand, and F. Chaumette used SIFT features to implement automatic initialization. Generally, the SIFT features is 128-dimensional, and the traditional matching algorithm is larger and more complicated.The paper based on the virtual visual servoing algorithm, matches the features by using KD-tree-based nearest neighbor search algorithm based on SIFT. And then using the linear algorithm calculates the initial position of the camera. Thereby, the automatic initialization is completed. This method reduces the time complexity, to some extent, improves the original algorithm, and is more efficient and convenient. However, the improved algorithm is still unable to satisfy the needs of real time tracking, In order to closer to the needs of real time tracking, the paper uses the randomized trees as the classification technique. Then, treat the feature point matching as a classification problem. It further reduces the automatic initialization time and enhances the efficiency of the algorithm.